Posted by Duke [ 208.98.184.1 ] on November 06, 2007 at 15:42:52:
In Reply to: crossover posted by valveman on November 01, 2007 at 13:47:30:
Hello valveman,
First off, my apologies for how long it took me to reply. I have been busy getting some speakers ready for delivery and haven't checked in here as much as I normally do, so didn't realize you had asked me some questions.
The crossover topology I use looks like third order filters both highpass and lowpass, but with unusual component values. The acoustic slopes are gentler than third order in the crossover region and then accelerate to a lot steeper than that once we're about a half-octave away from the crossover region on either side. I do not set out with a given crossover philosophy carved in stone; other areas are much higher in priority for me so I use a crossover topology that gives the results I want in those other areas. In the case of Stormbringer, the Jazz Modules, and the Dream Makers, a third-order-like topology has worked well.
When desiging a crossover, the first thing I do is take an impedance curve of each driver, with both drivers mounted in the cabinet. Then I take a series of frequency response measurements at different angles and distances, with and without time-gating depending on what I'm focusing on.
Next I use a computer modelling program to rough out the crossover, based on the data I've gathered. I then build & measure the crossover, which sits on the floor behind the speaker. The crossover is never right the first time, so I use the computer program to save time in modelling different possible fixes for whatever's not right. Then I modify or rebuild the crossover and measure again. Sometimes the drivers I'm using simply will not do what I want done, so I have to get new drivers and start over. I went through over a dozen different compression drivers and over a half-dozen different woofers in designing Stormbringer, for instance.
Finally it measures pretty good, so I listen to several different tracks. Single-speaker listening is better for hearing frequency response anomalies, as you aren't distracted by imaging. I listen to see if something doesn't sound right, and when I hear it (inevitable early in the process) I try to figure out what the problem is. This is the hard part - the data looks pretty good at a glance, but something isn't quite right. I have to find it so that I can fix it without screwing something else up at the same time. This usually involves looking more closely at the data and/or taking additional measurements.
Now all this time I am playing not only with frequency response, but also with the impredance curve. I want both to fall within the boundaries I have in mind, and adding the impedance curve requirement to the mix roughly quadruples my workload.
Towards the end of the process there will be some judgement calls to make in component values, where it sounds good both ways. At this stage I do a combination of listening and measuring, but I'm wary that going by ears alone as that can lead me astray. My ears are fallible for fine-tuning; they're much better at telling me that there's an audible problem than they are in picking between two equally good-sounding (to me at least) alternatives.
I don't hold the crossover board on my lap and tweak by ear. All of my tweaking is by measurements, then checked by ear. Sometimes I can quickly and reliably hear a difference when I make a change and sometimes I can't. Sometimes I can hear a difference and have a preference when I make a change, and sometimes not. On the other hand the test instruments are very good at not only detecting the difference (if any - sometimes there is none) but also telling me whether or not the change is in the right direction.
My approach implies a lot of trust in measurements, and I can get away with that because of the suite of measurements that I take. It is quite time-consuming but that's what it takes to get a complete enough picture. Problems with measurements arise when the measurement suite does not provide enough data and/or when psychoacoustic considerations are ignored. Psychoacoustics is what I use to decide how much weight to give to each the various measurements I take.
Hope this helps. Let me know if you have any additional questions or would like clarification.
Duke
[ AudioKinesis Forum ] [ Help ]